One hundred thirty adolescent females (ages 12- 16) referred from the juvenile justice system will be randomly assigned to one of two treatment conditions: Treatment Foster Care (TFC) or Group Care (GC). All participating girls will have been screened and recommended for placement in out-of-home care. In TFC one girl is placed in a family home where the foster parents have been recruited and trained, and are supervised to provide a set of treatment components that are hypothesized to be related to specific short- and long-term outcomes. In GC, girls are placed with from 6 to 13 peers who are experiencing similar problems with delinquency. The purpose of the study is twofold: to systematically evaluate the short-and long-term outcomes of subjects participating in the experimental intervention (TFC) relative to those in the control condition, and to evaluate the contribution of the treatment components to immediate and longer-term outcomes. At the group level, we expect that girls in TFC relative to those in GC will, in the short run, experience less association with antisocial peers; fewer negative departures from treatment, more contact with prosocial peers, better school adjustment and performance, and fewer high-risk sexual contacts. Long-term, we expect girls in TFC relative to those in GC to have less delinquency and drug use, better relationships with non-antisocial romantic partners, higher rates of school completions, better occupational functioning, and fewer early pregnancies and sexually transmitted diseases. At the individual level, we hypothesize that to the extent that girls receive key treatment components (i.e., close supervision, consistent discipline, negotiation/anger expression skills, monitoring and help with school work, education on avoiding high-risk sexual contacts) they will have better short- and long-term outcomes regardless of placement, condition. Further, we expect that to the extent that the girl's post-treatment living situation supports these components, she will do well in follow- up. Multimethod/multiagent measures will be used to form indicators of constructs of treatment components and outcomes. MANOVAs will be used to test for group-level effects and multiple regressions analyses will be used to examine the contribution of treatment components an characteristics of living situation to individual-level outcomes.